Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
4460
~
44
65
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
5
.
pp4460
-
44
65
4460
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Recomm
ender s
ystem f
or p
er
s
onali
sed trav
el itine
ra
ry
Tanu
ja Cho
u
dha
r
y
B
, Tula
si B
Depa
rtment
o
f
C
om
pute
r
Scie
n
ce,
CHRIS
T
(De
e
m
ed
to
b
e
Unive
rsit
y
)
,
Ind
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
23
, 201
9
Re
vised
A
pr
17
, 2
01
9
Accepte
d
Apr
26
, 201
9
A
rec
om
m
ende
r
s
y
stem
is
an
ap
proa
ch
to
giv
e
a
n
appr
opri
at
e
soluti
on
to
a
par
ticula
r
problem
.
Thi
s
hel
ps
i
n
rec
ognising
th
e
patter
n
or
beh
avi
our
of
a
user
to
suggest
future
poss
ibl
e
l
ike
s
of
the
user
.
Now
aday
s
peo
ple
li
k
e
to
tra
ve
l
during
their
spare
ti
m
e,
it
has
bec
om
e
a
ri
gid
ta
sk
to
deci
de
where
to
go.
Thi
s
pape
r
r
epr
ese
nts
a
customised
rec
om
me
nder
s
y
st
em
to
hel
p
users
i
n
desti
ning
the
ir
i
ti
ner
ar
y
.
A
m
od
el
is
d
esigne
d
t
o
suggest
the
b
e
st
pla
c
es
to
visit
in
Rom
e.
A
questi
onnai
r
e
was
pre
par
ed
to
g
et
informa
ti
on
a
bout
user’s
int
er
est
during
t
hei
r
tra
v
el
.
Th
e
m
odel
gene
rates
the
best
five
pla
ce
s
to
visi
t
with
respe
ct
to
t
he
cho
ic
e
pi
cked
b
y
th
e
user
.
T
he
top
five
pl
aces
for
ea
ch
ca
t
egor
y
will
b
e
display
ed
to
t
he
user
and
the
user
was
aske
d
to
pic
k
a
start
ing
p
oin
t
for
the
it
in
era
r
y
.
The
n
the
m
odel
gene
rates
anot
h
er
set
off
a
fil
tered
li
st of
p
l
ac
es
to
enh
ance
t
hei
r
tra
v
el
expe
r
ie
nc
e.
It inc
lud
es
display
ing
the
top
5
resta
u
r
ant
s to
visit
duri
ng
their
t
rav
e
l.
Ke
yw
or
d
s
:
Re
com
m
end
er
syst
e
m
To
urped
ia
Trav
el
To
ur
ism
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Tan
uj
a
Ch
oudhary B,
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nce,
CHRIST
(Dee
m
ed
to
be Un
i
ver
sit
y),
No
-
17, Mee
ra
Bhava
n, 2n
d
cr
os
s,
Mu
nn
i
kr
is
hn
a
ppa
Lay
ou
t
, Ad
ugodi, Be
ngal
uru
56
0030,
Ind
ia
.
Em
a
il
:
ta
nu
j
ac
houdha
ry.c
houdha
ry@
gm
ai
l.
com
1.
INTROD
U
CTION
A
rec
omm
end
er
syst
em
is
a
syst
e
m
us
ed
f
or
i
nfor
m
at
ion
filt
ering.
T
he
syst
e
m
le
arn
s
inf
or
m
at
ion
from
the
past
data
to
gi
ve
a
su
ggest
io
n.
It
is
a
too
l
use
d
to
a
bs
or
b
detai
l
of
in
for
m
at
ion
in
a
s
pecific
env
i
ronm
ent.
Each
e
nviro
nm
ent
is
a
f
or
m
of
di
ff
e
ren
t
dom
ai
ns
su
ch
a
s
e
-
c
omm
erce,
tourism
,
so
ci
al
-
m
edia,
adv
e
rtise
m
ent
and
et
c.
W
it
h
the
ex
plosi
ve
gro
wth
of
soc
ia
l
m
edia
and
de
vel
op
m
ent
of
web
2.0
,
la
rge
a
m
ou
nts
of
tra
vel
in
form
at
io
n
a
re
bein
g
up
loade
d
per
m
inu
te
on
trav
el
web
sit
es
[
1].
I
n
a
dom
ai
n,
th
ere
is
a
v
ari
ou
s
num
ber
of
fact
ors
w
hi
ch
can
a
ff
ect
i
n
the
way
the
r
ecom
m
end
er
m
od
el
beh
a
ves
.
O
ne
s
uch
nee
d
wa
s
in
the
fiel
d
of
e
-
com
m
erce,
du
e
to
the
incre
ase
of
gro
wth
in
nu
m
ber
of
c
us
tom
ers
durin
g
the
la
st
fo
ur
ye
ars
on
ly
a
bout
15
%
(
2010
-
2014)
[
2].
T
his
le
d
t
o
a
sig
nificant
need
to
buil
d
a
rec
omm
end
er
fr
am
ewo
r
k
t
o
choose
wh
ic
h fact
ors
yi
el
d
an
acc
ur
a
te
o
utc
om
e.
Re
com
m
end
er
syst
e
m
s
are
us
ually
cl
assifi
ed
acco
rd
i
ng
t
o
their
ap
proac
h
to
rati
ng
est
im
at
ion
[3
]
.
This
syst
em
is
cl
assifi
ed
int
o
the
f
ollow
i
ng
ty
pes,
Coll
a
borati
ve
filt
erin
g,
C
on
te
nt
-
ba
se
d,
Kno
wled
ge
base
d
and
Hyb
rid
re
com
m
end
ers
.
Coll
aborati
ve
f
il
te
ring
is
pro
ba
bly
the
m
os
t
fam
i
li
ar,
m
os
t
widely
im
ple
mente
d
and
m
os
t
m
a
ture
of
the
te
c
hnologies
[
4].
It
ta
kes
into
c
ons
iderati
on
the
vi
ews/rati
ngs
of
oth
e
r
pe
ople
wh
e
n
decidin
g
on
r
ecom
m
end
at
io
ns
;
so
m
et
i
m
es
it
s
nar
rowed
down
int
o
a
sp
eci
fic
dem
ogra
ph
ic
with
si
m
il
ar
interest
s.
This
filt
ering
has
so
m
e
chall
eng
e
s
nam
el
y
data
sp
ar
sit
y
and
scal
a
bili
t
y
[5
]
.
Con
te
nt
-
bas
e
d
reco
m
m
end
ers
m
ake
decisi
on
s
base
d
on
w
hat
the
us
er
ha
s
pr
e
viously
rated
or
w
hat
the
us
er
is
cu
r
ren
tl
y
lookin
g
at
[
6].
Kno
wled
ge
-
ba
sed
set
out
a
s
uggestio
n
on
t
he
basis
of
us
e
r’
s
nee
ds
an
d
t
ast
es.
It
m
at
ch
es
the
need
of
t
he
pa
r
ti
cular
it
e
m
wit
h
the
nee
d
of
a
par
ti
cular
us
e
r
to
giv
e
a
possi
ble
sugg
e
sti
on.
T
he
hy
br
id
sy
ste
m
is
a
com
bin
at
ion
of
Co
ntent
-
ba
sed
an
d
Coll
a
borati
ve
filt
eri
ng.
W
it
h
the
he
lp
of
histo
rical
inform
ation
and
th
e
new in
form
at
io
n of t
he use
r
is
consi
der
e
d
a
nd
an
al
yse
d t
o gi
ve
the
r
ec
omm
end
at
io
n.
This
pap
e
r
f
oc
us
es
on
the
fie
ld
of
to
ur
ism
.
As
a
m
ajo
r
pa
r
t
of
t
he
m
od
er
n
se
rv
ic
e
i
ndust
ry,
tourism
has
ex
per
ie
nce
d
rap
i
d
gro
wth
ov
er
the
pa
st
decad
e
[
7].
Th
e
tou
rists
of
to
day
are
ver
y
dem
and
in
g
an
d
have
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
mm
e
nder
syste
m
fo
r
pers
onalised
travel
it
inerary
(
T
an
uja
C
houd
ha
ry
B
)
4461
com
plex,
m
ult
i
-
la
ye
red
desir
es
an
d
nee
ds
[
8].
T
rav
e
l
a
nd
tourism
is
an
area
w
her
e
th
e
m
ajo
rity
of
people
com
e
un
de
r.
T
her
e
a
re
m
any
facto
rs
wh
ic
h
aff
ect
to
buil
d
a
n
ef
fici
ent
reco
m
m
end
er
syst
e
m
.
Du
e
t
o
this
choosi
ng
relat
ively
,
i
m
po
rtan
t
factor
s
is
diffi
cult.
In
orde
r
to
searc
h
an
d
re
com
m
end
touri
st
sp
ots
eff
ect
i
vely
,
it
is
first
neces
sary
to
cha
ract
erise
the
to
ur
is
t
spots
[
9].
Ma
ny
use
rs
prefe
r
a
need
f
or
pe
r
so
na
l
rec
omm
e
nd
e
r
syst
e
m
w
hich ca
n
gi
ve
them
a set o
f
a
n
it
ine
rar
y o
n
the
b
as
is of
thei
r
pref
eren
ces
on tra
ve
li
ng
. T
o
a
ddre
ss this
issue, I
n
this
pa
pe
r
, a recom
m
end
er
syst
em
is prop
os
e
d
to
enh
a
nce
us
e
rs
t
rav
el
e
xperie
nc
e.
2.
LIT
ERATUR
E REVIE
W
Durin
g
le
isu
re
tim
e
us
ers
pr
e
f
er
to
t
rav
el
.
Tr
avel
any
wh
e
re
arou
nd
the
w
or
l
d.
When
a
use
r
decides
on
wh
e
re
to
vi
sit
i
t
al
ways
b
ecom
es
a
te
dio
us
ta
sk
to
deci
de
w
her
e
to
go
first.
Nowa
da
ys
there
are
m
ulti
ple
op
ti
ons
a
vaila
ble
w
hich
he
lp
the
us
e
r
ov
e
rc
om
e
this
ta
sk
.
A
rec
omm
end
er
sy
stem
is
a
so
luti
on
.
Re
com
m
end
er
syst
e
m
helps
end
us
e
rs
to
narrow
do
wn
on
wh
ic
h
places
t
o
visit
.
I
n
[10]
a
par
ti
cula
r
lo
cat
io
n
was
exam
ined
u
sin
g
a
crow
ds
ou
rcin
g
ap
proach.
The
m
ain
f
ocu
s
was
on
the
am
ou
nt
of
cr
owd
pr
es
ent
at
a
par
ti
cula
r
locat
ion
.
The
pro
po
sed
m
od
el
not
on
ly
c
on
si
ders
inf
or
m
at
ion
from
the
blo
gs,
we
b
pag
es
,
s
ens
or
read
i
ng
s
,
et
c…
bu
t
al
so
ta
kes
in
us
e
rs
vi
ewpoint.
W
it
h
the
help
o
f
TSRS
(Tour
i
st
Sp
ot
Re
co
m
m
end
er
Syst
e
m
)
and
s
et
of
locat
ion
su
ggest
e
d
to
the
us
e
rs
the
sy
stem
per
form
s
j
us
t
-
in
-
ti
m
e
inform
ation
en
richm
ent
for
tho
s
e
sel
ect
ed
set
of
locat
ion
.
T
he
pro
po
sed
syst
em
is
desig
ne
d
for
m
ob
il
e
us
ers
to
en
han
ce
thei
r
travel
exp
e
rie
nce
.
T
he
gro
wth
of
i
nfor
m
at
ion
is
inc
reasin
g
dr
ast
ic
al
ly
wh
ic
h
le
ad
s
to
t
he
c
oncer
n
of
bi
g
data. D
ue
t
o
vast
c
halle
nge
s
in
the
fiel
d
of
bi
g
data
[11],
reco
m
m
end
er
syst
em
is
so
luti
on
f
or
s
om
e.
The
pa
pe
r
us
es
Singular
value
deco
m
po
sit
io
n
(SV
D)
to
outper
form
the
chall
eng
es
face
d
in
the
fiel
d
of
big
data.
SVD
base
d
reco
m
m
end
er
syst
e
m
is
pr
opose
d
to
eval
uate
and
le
ar
n
the
per
f
or
m
ance
an
d
accur
acy
in
a
distr
ibu
te
d
env
i
ronm
ent. H
ad
oop an
d sp
ark are t
he
t
oo
l
s u
se
d
t
o per
form
the ex
pe
rim
ent.
So
m
et
i
m
es
the
te
xtu
al
de
ta
il
s
are
j
us
t
not
enou
gh
to
ca
ptu
re
c
om
plete
i
nfor
m
at
ion
ab
ou
t
a
re
ntal
place
hen
ce
th
is
pap
er
[
12]
con
si
ders
i
m
age
detai
ls
as
wel
l.
I
m
age
and
te
xtu
al
inf
or
m
at
ion
are
capt
ured
to
gen
e
rate
a
rec
omm
end
er
sys
tem
.
Cho
os
i
ng
a
vacati
on
re
ntal
was
a
c
ha
ll
eng
in
g
ta
sk.
This
pa
pe
r
us
e
s
thre
e
m
et
ho
ds
im
ages
-
ba
sed
c
os
in
e
si
m
i
la
rity
ca
l
culat
ion
,
te
xtua
l
descr
ipti
on
-
base
d
Jacca
r
d
si
m
il
arity
cal
c
ulati
on
and
t
he
f
us
io
n
of
both
the
m
et
ho
ds
.
A
hybri
d
rec
omm
end
er
syst
e
m
was
bu
il
t
to
sug
gest
pre
fer
a
ble
accom
m
od
at
ion
to
the
us
e
r.
In
sp
it
e
of
t
he
fact
that
m
any
travel
propo
sal
fr
am
ewo
r
ks
are
w
orked
with
a
m
ob
il
e
so
ci
al
netw
ork,
the
a
rtic
le
s
and
co
nvey
ed
m
essages
betwee
n
use
rs
are
sti
ll
not
us
ed
well
to
assess
us
ers
’
incli
nati
on
s
.
T
o
el
i
m
i
nate
this
issue
in
[
13
]
pa
pe
r
cl
oud
-
base
d
r
ecom
m
end
er
s
yst
e
m
is
pr
opos
e
d.
The
ide
ology
was
to c
om
bin
e the lat
est
clou
d
te
c
hnology,
an
al
yse
posted
b
lo
gs
on
t
he m
ob
il
e n
et
wo
r
k
syst
em
and
ide
ntify
use
rs’
patte
r
n
th
r
ough
se
nsors.
The
process
w
as
carrie
d
dow
n
in
t
hr
ee
dif
f
eren
t
sta
ges
na
m
ely
pre
-
to
ur
i
ng,
in
-
to
ur
in
g
a
nd
post
-
to
uri
ng.
E
ach
to
ur
f
un
ct
ion
s
a
re
diff
e
r
ent
from
the
oth
e
r.
T
he
pro
po
s
ed
syst
e
m
initial
l
y
con
str
ucts
m
et
a
-
gr
ou
p
ba
sed
on
sim
il
ar
us
er
prefe
rence
s.
F
ur
the
rm
or
e
,
with
t
he
help
of
CLOPE
al
gorit
hm
m
et
a
-
gr
ou
p
was
cl
assi
fied.
The
res
ult
wa
s
well
exam
ined
a
nd
s
uitable
P
OI
s
wa
s
giv
e
n
t
o
the users
dur
i
ng to
uri
ng.
Ther
e
are
dif
fe
ren
t
c
halle
nges
face
d
to
rec
om
m
end
travel
i
ti
ner
ary
to
a
use
r
or
gro
up
of
us
ers
ba
sed
on
thei
r
tourist
interest
and
c
ho
ic
e.
T
his
pa
per
[
14]
bu
il
ds
a
reco
m
m
end
er
m
od
el
t
o
overco
m
e
the
chall
eng
es.
The
pr
opos
e
d
al
gorithm
in
th
e
pap
e
r
co
ns
tr
uct
tourist
s
pa
st
PO
I
visit
based
on
their
ge
o
-
ta
gg
e
d
ph
otos
an
d
then
co
ns
tr
uct
a
m
od
el
of
use
r’
s
ch
oice
ba
sed
on
their
ti
m
e
sp
ent
visit
i
ng
eac
h
POI.
The
pro
posed
m
od
el
out
-
pe
rfor
m
s
va
rio
us
fact
or
s
t
o
gi
ve
a
relat
ively
bette
r
rec
omm
end
at
io
n
to
us
e
rs
base
d
on
t
heir
pr
e
fere
nces.
In
rece
nt
ye
ars
the
us
e
of
co
nt
extual
da
ta
is
beco
m
e
m
or
e
powe
rful
to
gi
ve
a
rec
omm
e
nd
at
io
n.
W
it
h
t
he
help
of
t
his
data
a
nd
co
ns
i
der
i
ng
t
he
pa
st
inf
or
m
at
ion
the
fo
ll
owin
g
[
15]
pa
pe
r
co
ns
tr
ucts
a
r
ecom
m
end
er
s
yst
e
m
.
The
rec
omm
end
er
f
ram
ewo
r
k
gets
l
og
ic
al
da
ta
by
m
ining
cl
ie
nt
rev
ie
ws
and
joi
ning
the
m
with
cl
ie
nt
r
at
ing
histor
y
t
o
fi
gur
e
a
util
it
y
capaci
ty
ov
er
a
n
a
r
rangem
ent
of
t
hings.
I
n
the
fra
m
ewo
r
k,
t
he
set
ti
ng
de
r
ivati
on
is
disp
la
ye
d
as
a
regulat
ed
s
ub
je
ct
dem
on
strat
ing
iss
ue
i
n
w
hich
a
n
a
rr
a
ng
e
m
ent
of
cl
ass
es
f
or
a
rele
va
nt
trai
t
est
ablishes
the
po
int
set
.
Wor
ds
play
an
im
p
or
ta
nt
ro
le
in
the
com
m
ent
on
a
place
or
a
n
obj
ect
.
The
se
words
can
be
us
e
d
a
nd
hel
ped
i
n
ge
ner
at
in
g
a
re
com
m
end
er
syst
e
m
.
This
pa
per
[16]
ge
ne
r
at
es
tem
po
ral
feature
vecto
rs
on
pas
sing
ob
j
ect
s
to
their
pro
pose
d
reco
m
m
end
er
syst
em
.
The
syst
e
m
evaluates
by
i
den
ti
fyi
ng
the
vo
ca
bula
ry
r
el
at
ed
to
the
obje
ct
s
with
t
he
he
lp
of
Wiki
pedi
a,
fin
ding
t
he
tre
nd
of
al
l
t
he
obj
ect
s
with
the
hel
p
of
Twitt
er
a
nd
to
feat
ur
e
t
he
he
aviness
of
word
s
co
ntaine
d
i
n
eac
h
disti
ngui
sh
ed
t
rend
t
o
get
te
m
po
ral
fe
at
ur
e
vecto
rs
for
ea
ch
obj
ect
.
The
res
ultant
vect
or
s
preci
sel
y
r
et
urn
the
rese
m
blance
of
a
po
i
nt
of
i
ntere
sts
for
a
ppoin
te
d
ti
m
e
dur
at
io
n.
Re
com
m
end
er
syst
e
m
is
hel
pful
to
giv
e
re
com
m
end
at
io
n
on
bases
of
hi
storical
data
bu
t
wh
e
n
it
com
es
to
new
us
ers
it
fail
s
to
per
f
orm
at
the
m
axi
m
u
m
.
T
o
res
olv
e
this
i
ssu
e
the
pap
e
r
of
[
17]
propo
sed
a
go
al
-
base
d
a
ppr
oac
h.
T
he
pap
e
r
us
es
c
onte
nt
base
d
fi
lt
ering
a
nd
c
oll
aborati
ve
fi
lt
ering
us
in
g
K
-
nn.
The
c
om
bin
at
ion
of
the
t
wo
te
chn
iq
ue
t
o
gi
ve
ap
pro
pr
ia
t
e
an
d
rele
van
t
ov
e
rc
om
e
to
t
he
ne
w
us
e
r
pro
file
.
Go
al
base
d
a
ppr
oac
h
with
K
-
nn,
it
c
om
par
es
the
sim
il
ari
ti
es
betwee
n
use
rs
to
gi
ve
s
uggestio
n.
It
was
a
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
4
4
6
0
-
4
4
6
5
4462
cru
ci
al
ta
s
k
to
colle
ct
ap
pro
pr
ia
te
data
f
rom
a
vast
a
m
ou
nt
of
in
form
at
ion
for
a
re
com
m
end
er
syst
e
m
.
The
pa
per
of
[18]
propose
d
the
arch
it
ect
ur
e
of
to
ur
ist
su
pport
inf
orm
at
ion
syst
e
m
wh
ic
h
incl
udes
VR
con
te
nts
to
prom
ote
Iw
at
e
a
rea
in
J
apa
n
and
gathers
co
nt
ent
reposit
or
y
and
trai
ning
data
to
bu
il
d
r
egio
nal
sp
eci
fic
rec
omm
end
er
e
ngine
on
t
he
to
ur
ist
support
syst
e
m
.
The
author
s
in
[
19]
util
ise
te
xtu
al
in
for
m
at
ion
from
us
ers
revi
ews
an
d
us
e
ad
-
hoc
an
d
re
gr
essi
on
-
base
d
reco
m
m
end
at
i
on
m
easur
e
s
to
giv
e
a
pe
rs
onal
ise
d
rati
ng
val
ue.
T
he
m
od
el
pr
oves
that
the
acq
uire
d
rati
ng
va
lue
gi
ves
a
be
tt
er
sco
re
tha
n
the
der
i
ved
r
at
in
g
value
giv
e
n
by
the
us
e
rs.
T
he
work
i
n
[
20]
est
ablishes
a
co
nnect
ion
w
it
h
t
he
us
er b
y
colle
ct
ing
in
form
at
i
on
o
f
us
ers
intere
s
ts
and use
s
his/he
r
c
urren
t c
onte
xt to rec
omm
e
nd n
ea
r
by spot
s b
ase
d o
n
thei
r
intere
st.
3.
RESEA
R
CH MET
HO
D
Vast
am
ou
nt
of
in
form
at
ion
is
on
the
i
nter
ne
t,
inf
or
m
at
ion
ab
ou
t
places
t
o
visit
,
restau
r
ants,
m
ov
ie
s,
sh
op
ping
ce
nter,
hote
ls
an
d
m
any
m
or
e.
It
has
be
c
om
e
a
necessit
y
to
go
t
o
the
to
p
place.
T
his
m
akes
it
diff
ic
ult
f
or
use
rs
to
de
ci
de
or
pic
k
am
on
g
the
vast
op
ti
ons.
To
de
al
wi
th
this
iss
ue
a
syst
e
m
is
pro
pos
e
d
nam
ed
reco
m
m
end
er
syst
em
.
The
reco
m
m
end
e
r
syst
e
m
i
s
a
te
chn
iqu
e
t
o
filt
er
ou
t
the
inform
ation
and
giv
e
su
ggest
io
n
t
o
e
nd u
s
ers. F
ollo
wing a
re th
e
steps t
o bu
il
d a r
ecom
m
end
er
s
yst
e
m
:
Step
1: G
at
her
i
ng d
at
a
Step
2: P
reproc
ess and t
ran
s
f
or
m
d
at
a
Step
3: Build
a
m
od
el
Step
4: E
valuat
e the m
od
el
a
nd
gen
e
rate t
he
ou
tc
om
e.
R
too
l
is
use
d
to
ge
ne
rate
th
e
m
od
el
.
T
his
t
oo
l
pro
vid
e
s
a
set
of
sta
ti
sti
cal
an
d
gr
a
phic
al
te
ch
niques.
It
inclu
des
re
gr
essi
on,
sta
ti
sti
cal
infer
e
nc
e,
m
achine
le
arn
i
ng
al
gorit
h
m
,
data
anal
ysi
s
to
nam
e
a
fe
w.
R
pr
ovide
s
va
rio
us
pac
kag
e
thr
ough
w
hich
var
io
us
ope
ra
ti
on
can
be
pe
rfor
m
ed
on
th
e
data.
So
m
e
of
the
pack
a
ges
use
d
to
buil
d
the
rec
omm
end
ed
m
od
el
are
google
way,
dp
y
r,
Im
a
p
to
nam
e
few
.
For
i
ns
ta
nce
,
I
m
ap
pack
a
ge
hel
ps
in
c
om
pu
ti
ng
the
geo
dista
nce
betwee
n
t
wo
points
sp
e
ci
fied
by
la
ti
t
ud
e/
lo
ngit
ude
us
in
g
Vince
nty
inv
e
r
se
f
or
m
ula
for
el
li
ps
oid
s
.
He
nce,
it
is
a
po
werfu
l
t
oo
l
t
o
get
ap
pro
pr
ia
te
insig
ht
int
o
th
e
raw
data an
d hel
ps
in d
eci
si
on m
a
king.
3.1. D
atase
t
The
data
is
co
ll
ect
ed
fr
om
ww
w
.to
ur
-
pe
dia.
org
.
T
he
li
nk
consi
sts
of
dat
a
on
8
di
ff
e
re
nt
Euro
pean
ci
ti
es.
Each
ci
ty
is
cat
ego
rie
s
with
fou
r
ty
pes
nam
ely
Acco
m
m
od
at
ion,
Re
sta
ur
a
nt,
Po
int
of
I
ntere
st
and
Attract
ion.
In
t
his
pa
per
P
oin
t
of
I
nterest
an
d
Attract
ion
of
Rom
e
is
con
sidere
d.
Data
fro
m
tou
rp
e
di
a
co
ns
ist
s
of
12
at
tri
bu
te
s
su
c
h
a
s
nam
e,
id,
Ca
te
gory
,
locat
io
n,
la
ti
t
ud
e
an
d
l
ongit
ud
e
,
po
la
rity
,
address
,
s
ub
Ca
te
gory,
or
i
gin
al
I
d,
deta
il
s,
and
re
view
.
W
it
h
res
pect
to
the
a
bove
at
tribu
te
s,
the
da
ta
set
con
sist
s
of
28
198
obser
va
ti
on
in
total
.
The
de
ta
il
s
of
eac
h
place
in
Rom
e
is
colle
ct
ed
usi
ng
the
G
oogl
e
Plac
es
API.
This
API
is
i
ni
ti
at
ed
us
in
g
the
R
too
l,
in
wh
ic
h
google
way
pac
kag
e
is
us
e
d.
This
pac
kag
e
has
the
G
oogl
e
Plac
es
AP
I
m
et
ho
d,
wh
ic
h
is
us
ed
to
retrieve
the
place
detai
ls.
The
data
is
colle
ct
ed
f
ro
m
end
us
e
r’
s
pe
rs
pe
ct
ive
on
the
ba
sis
of
their
pr
e
fer
e
nc
es an
d
c
hoic
e f
or the tra
vel.
3.2. Pre
-
proc
essing
Data
pre
proces
sing
is
the
proc
ess
of
tra
nsfo
r
m
ing
data
into
a
f
or
m
that
is
f
it
fo
r
an
al
ysi
s.
This
is
due
to
the
var
i
ou
s
facto
rs
li
ke
da
ta
no
t
bei
ng
in
the
rig
ht
da
ta
ty
pe,
unstr
uc
tur
e
d
f
or
m
at
,
erron
e
ous
data,
et
c.
Ther
e
a
re
sev
eral
te
chn
iq
ue
s
by
wh
ic
h
da
ta
can
be
transfo
rm
ed
int
o
the
rig
ht
f
or
m
at
fo
r
ana
ly
sis.
Data
-
m
ining
ha
s
te
ch
niques
t
o
pr
e
process
th
e
data
a
nd
it
al
so
has
a
colle
ct
ion
of
var
io
us
te
chn
i
qu
e
s
to
e
xtract
patte
rn
s
a
nd
to
buil
d
m
od
el
s
f
ro
m
la
rg
e
data
-
set
s
[
21]
.
T
he
data
c
ollec
te
d
for
this
resea
rc
h
had
er
r
on
e
ou
s
an
d
unwa
nted data
wh
ic
h had t
o b
e filt
ered i
n order
to
c
reate t
he
r
ec
omm
end
e
r
syst
em
.
3.2.1. D
ata
c
le
an
in
g
The
dataset
f
r
om
tou
rpedia
consi
sts
of
12
at
tribu
te
s
s
uc
h
as
id,
nam
e,
address,
ori
gi
na
lId,
la
ti
tu
de,
longit
ud
e
,
loca
ti
on
,
s
ubCa
te
gory,
Ca
te
gory
po
la
rity
,
detai
ls,
an
d
rev
ie
ws.
The
nam
e
colum
n
had
obser
vations
wh
ic
h
ha
d
garbag
e
values
w
hich
ha
d
to
be
rem
ov
ed.
T
his
was
done
us
in
g
R.
A
functi
on
was
w
ritt
en
to
che
c
k
and o
m
it
the observati
ons
w
hi
ch
di
d no
t
ha
ve
ch
a
racters
from
a
-
z an
d 0
-
9 i
n
the
n
am
e co
lum
n.
3.2.2. D
ata
re
duct
i
on
Af
te
r
da
ta
cl
eanin
g
was
pe
rfo
rm
ed
the
data
was
furthe
r
re
du
ce
d
to
c
onsi
der
only
the
na
m
e
colum
n.
The
place
nam
es
wer
e
the
n
pa
ssed
to
G
oogl
e
Plac
es
AP
I
to
fetch
the
det
ai
ls
of
each
pl
ace.
Fo
r
eac
h
place
-
nam
e
rati
ng
a
nd
ty
pes
we
re
c
on
si
der
e
d.
T
he
colum
ns
wh
ic
h
had
m
ulti
ple
rati
ng
val
ues
wer
e
the
n
re
du
ced
by
consi
der
i
ng
th
e
ave
rag
e
rati
ng.
Am
on
g
the
val
ue
fetche
d
t
her
e
we
re
few
colum
ns
w
hich
ha
d
rati
ng
val
ue
ha
s
0,
t
hese
obser
va
ti
on
s
wer
e
o
m
it
te
d
from
the d
at
aset
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
mm
e
nder
syste
m
fo
r
pers
onalised
travel
it
inerary
(
T
an
uja
C
houd
ha
ry
B
)
4463
4.
IMPLEME
N
TATION
This
sect
ion
pro
vid
es
detai
ls
about
the
pro
po
s
ed
rec
omm
end
e
r
syst
em
.
The
syst
em
gen
erates
the
top
5
places
t
o
visit
durin
g
their
sta
y
in
R
om
e.
The
s
uggestio
n
of
pla
ces
was
bas
ed
on
the
an
swe
r
to
t
he
cat
egory
of
the
quest
io
ns
aske
d
to
us
er
.
The
quest
io
nn
ai
re
co
ntain
s
a
set
of
pr
edef
i
ned
ques
ti
on
s
.
These
quest
io
ns
a
re
to
be
a
ns
we
re
d
by
use
rs
based
on
t
heir
c
hoic
e
an
d
pr
e
fer
e
nces
.
Am
on
g
them
thre
e
qu
e
sti
on
s
hav
e
one
or
m
or
e
s
ub
quest
io
n
an
d
t
hr
ee
quest
io
ns
ha
ve
no
s
ub
quest
io
n.
The
respo
ns
e
was
ta
ke
n
in
the
form
of
bin
ary
ei
ther
ye
s
or
no.
De
pendin
g
on
th
e
us
er
re
qu
i
re
m
ents
the
recom
m
end
er
syst
e
m
wi
l
l
gen
e
rate
out
put.
The
pro
pose
d
m
od
el
ta
kes
the
in
pu
t
from
the
us
e
r
a
nd
pr
ocesses
it
.
Fig
ur
e
1
c
onsist
of
so
m
e
of
t
he
quest
io
ns
as
ked
to
th
e
us
er
.
T
he
outpu
t
w
as
ge
ne
r
at
ed
on
the
ba
sis
of
past
database.
The
dat
abas
e
consi
sts o
f 133
6 ob
s
er
vations.
Each
sam
ple
was
diff
e
re
ntiat
ed
f
ro
m
the
ty
pe
value.
This
ty
pe
value
co
nsi
sts
of
17
different
le
ve
ls.
The
le
vels
ar
e
art_
galle
ry,
Historical
_pla
ce,
m
us
eum
,
Mon
um
ent,
m
os
que,
place
_of_wors
hip
,
syna
gogue,
Rom
an_
tem
ple,
nat
ur
al
_f
eat
ure,
park,
zo
o,
ca
m
pg
r
ound,
a
m
us
e
m
ent_p
ar
k,
book
_s
to
re,
li
br
ary,
l
od
ging
a
nd
pr
em
ise
.
Ther
e
are
22
obser
va
ti
on
with
rati
ng
5
f
or
a
rt_
galle
ry.
W
it
h
t
he
he
lp
of
sam
ple
fu
nctio
n
i
n
R
at
each
run
ra
ndom
fi
ve
places
are
disp
la
ye
d
to
the
us
e
rs.
Th
er
e
are
on
ly
14
ob
s
er
vations
unde
r
the
cat
eg
or
y
of
histor
ic
al
place
.
These
obse
rvat
ion
s
fall
unde
r
rati
ng
of
4.6
,
4.7
and
4.9.
At
each
run
an
y
five
rando
m
places
are
dis
play
ed.
W
it
h
res
pect
to
m
on
um
ent
t
her
e
a
re
12
ob
serv
at
io
ns
a
nd
few
sam
ples
are
distrib
uted
in
to
4.
7
and
a
nd
few
are
distrib
uted
into
4.6
rati
ng.
A
set
of
5
rand
om
places
are
disp
la
ye
d
fr
om
the
set
of
12
ob
s
er
vation.
In
total
there
are
136
obse
rv
at
i
on
with
the
ty
pe
value
m
us
e
um
.
The
data
was
distrib
ute
d
well
acro
s
s
rati
ng
r
ang
i
ng
from
5
to
2.
F
r
om
the
total
a
sing
le
sa
m
ple
is
ran
dom
ly
picked
from
the
fo
ll
ow
i
ng
rati
ng v
al
ue
5,
4.9, 4.8,
4.7 an
d 4.6.
Rom
e
is
fa
m
ou
s
f
or
it
s
ch
ur
c
hes
due
to
this
fact
there
are
596
obse
rv
at
i
on
unde
r
place
_of
_wo
rsh
i
p.
Am
on
g
t
hem
39
obser
vatio
ns
ha
ve
rati
ng
5
from
wh
ic
h
to
p
5
is
ra
ndom
l
y
picke
d.
U
nd
er
m
os
que
ty
pe
ther
e
are
6
obse
rv
at
ion
w
hich
fall
in
the
f
ollowi
ng
rati
ng
value
s
uch
as
4.7
,
4.5
,
4.2
a
nd
4.1.
From
these
set
,
rand
om
5
plac
es
of
m
os
qu
e
are
ge
ner
at
e
d
and
disp
la
ye
d
to
the
us
e
r.
Th
ere
are
only
3
synag
ogue
places
t
w
o
of
wh
ic
h
is
rated
4.4
a
nd
t
he
oth
e
r
is
rated
3.6.
All
the
thre
e
places
was
s
uggeste
d
to
the
us
ers
.
I
n
the
da
ta
set
there
are
only
two
places
of
the
ty
pe
r
om
an
tem
ple
on
e
of
wh
ic
h
is
4.7
and
t
he
ot
her
i
s
4.6.
B
oth
of
these
places
are
sug
gested
t
o
the
us
ers
on
t
heir
visit
only
if
they
are
inte
r
est
ed
in
visit
ing
r
om
an
temple
in
Rom
e.Fo
r
the
us
ers
who
are
natu
re
lov
e
rs
a
nd
inte
r
est
ed
in
visit
ing
local
par
ks
at
Ro
m
e
a
sing
le
sa
m
ple
is
rand
om
l
y
picked
f
ro
m
the
rati
ng
value
bet
w
een
4.4
to
4.8
of
natu
ral_f
e
at
ur
e
a
nd
4.3
to
4.7
of
pa
rk.
Th
ere
are
4
obse
rv
at
io
n
with
the
ty
pe
value
z
oo
of
wh
ic
h
th
ree
ha
ve
a
rati
ng
of
4.3
an
d
the
ot
her
h
as
a
rati
ng
of
2.1.
These
places
will
be
s
ugge
ste
d
to
the
use
rs
who
a
re
i
nterested
in
vi
sit
ing
zo
o
i
n
Rom
e.
Ther
e
are
8
ob
s
er
vations
w
hich
have
the
t
ype
valu
e
cam
pgr
ound
an
d
if
the
us
e
r
sel
ect
ye
s
from
these
five
rand
om
places
are
picke
d
a
nd
disp
la
ye
d t
o
t
he
u
se
r.
Ther
e
are
20
obser
vatio
ns
ha
ving
the
ty
pe
am
us
e
m
ent_p
ar
k,
am
on
g
thes
e
20
top
fi
ve
are
rand
om
l
y
picke
d
w
hich
hav
e
the
rati
ng
4
or
ab
ove.
F
or
th
os
e
use
rs
interest
ed
in
bookst
or
es
,
ther
e
are
11
ob
se
r
vations
from
wh
ic
h
the
top
fi
ve
places
are
ra
ndom
l
y
chosen
on
t
he
basis
of
rati
ng
value
w
her
e
ra
ti
ng
value
is
ei
ther
4
or
a
bove.
A
nd
la
stl
y,
fo
r
th
ose
us
ers
w
ho
w
ou
l
d
li
ke
to
vis
it
li
br
aries,
the
re
are
11
obser
vations
w
hich
hav
e
5
rati
ng
unde
r
li
br
a
ry
ty
pe,
fi
ve
places
are
picked
ra
ndom
ly
a
m
on
g
11
an
d
dis
play
ed
t
o
the
use
r.
Af
te
r
picking
the to
p place
s t
he user
w
a
s as
ked to
pick
a st
arti
ng point t
o st
art there
it
inerar
y, a
s s
how
n i
n
Fig
ure
2.
Figure
1. Q
ues
ti
on
nai
re
desig
n
Figure
2. Starti
ng point
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
4
4
6
0
-
4
4
6
5
4464
The
rec
omm
en
der
syst
em
wil
l
then
ge
ner
at
e
a
dist
ance
m
a
t
rix
has
s
how
n
in
Table
1.
Wi
th
the
help
of
dista
nce
m
a
trix,
the
ori
gina
l
li
st
of
places
to
visit
was
nar
r
ow
e
d
do
wn.
The
final
outc
om
e
wil
l
sh
ow
al
l
the
places
withi
n
8k
m
rad
iu
s
f
rom
the
sta
rting
po
i
nt.
Additi
onal
ly
,
top
5
re
sta
ur
a
nts
f
ro
m
the
sta
rt
in
g
place
is
su
ggest
e
d
to
th
e
us
er
.
G
oogle
Plac
es
API
m
e
thod,
wh
ic
h
is
us
e
d
to
retrie
ve
the
restau
ra
nt
detai
ls.
Ap
a
rt
from
the
quest
ions
aske
d
as
show
n
in
the
Fig
ur
e.
1,
pr
em
ise
and
l
odging
ar
e
so
m
e
add
it
ion
al
places
w
hi
ch
the
pro
po
se
d
syst
e
m
is g
oing to
r
ecom
m
end
. Fr
om
p
rem
ise
an
d
lo
dg
i
ng
t
op
f
ive p
la
ces is ra
ndom
ly
p
ic
ked w
hic
h
has rat
ing 4
or
above t
o
al
l t
he
u
se
rs.
Table
1.
Dista
nc
e m
a
trix
5.
RESU
LT
S
AND A
N
ALYSIS
The
f
ram
ewo
r
k
is
custom
ise
d
to
one
E
uro
pean
ci
ty
nam
ed
Rom
e.
This
gen
e
rates
the
resu
lt
of
the
places
nam
e
to
be
f
ro
m
Ro
m
e
.
The
ou
t
pu
t
of
the
m
od
el
give
s
a
set
of
s
pec
ific
place
nam
es
depen
ding
on
th
e
answer
to
a
spe
ci
fic
qu
est
io
n.
The
outc
om
e
was
so
le
ly
de
pe
nd
e
nt
on
us
e
r
s’
interest
an
d
cho
ic
es
.
The
re
su
lt
is
picke
d
f
r
om
the
past
c
ollec
ti
on
o
f
a
dataset
on
Rom
e.
Th
e
res
ult
shows
the
to
p
5
places
f
or
eac
h
qu
est
ion
aske
d
to
t
he
use
r.
T
hese
to
p
5
places
a
re
the
m
os
tl
y
pr
efera
bly
places
to
visit
durin
g
the
sta
y
in
Rom
e.
On
ce
t
he
use
r
picks
t
he
sta
r
ti
ng
point
of
t
he
jo
urney
al
l
filt
ered
li
st
of
pla
ces
as
s
how
n
in
Fig
ur
e
3
is
disp
la
ye
d.
T
hi
s
is
the
fi
nal
outc
om
e.
The
r
esult
s
hows
t
he
places
nam
e
within
8km
rad
ius
f
ro
m
the
sta
rting
po
i
nt
to
visit
duri
ng
t
heir
tra
ve
l.
Alon
g
wit
h
the
places
,
the
top
5
restau
rant
s
from
the
sta
rting
po
i
nt
was
al
so
reco
m
m
e
nd
ed t
o
the
us
e
rs
as
sh
ow
n
in
Fi
gur
e 4
.
Figure
3. To
p place
s
Figure
4. Nea
r
by
places
and
r
est
aur
a
nts
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Reco
mm
e
nder
syste
m
fo
r
pers
onalised
travel
it
inerary
(
T
an
uja
C
houd
ha
ry
B
)
4465
6.
CONCL
US
I
O
N
People
usual
ly
pr
e
fer
tra
velin
g
in
the
f
ree
ti
m
e
of
their
eve
ryday
li
fe.
It
ha
s
beco
m
e
a
ted
io
us
ta
sk
t
o
decide
on
t
he
it
inerar
y.
T
his
le
ads
to
the
ne
ed
of
a
rec
omm
end
er
syst
em
.
A
rec
omm
e
nd
e
r
syst
em
h
el
ps
in
choosi
ng
the top p
la
ces an
d
ge
ner
at
es a
n
ap
pro
pr
ia
te
o
utc
om
e o
n
wh
ic
h
pl
ace t
o
visit
. Th
ou
gh
the
re are
m
any
ty
pes
of
rec
omm
end
er
syst
e
m
s,
a
pe
rs
onal
ise
d
on
e
is
al
wa
ys
pr
e
ferred
.
A
custom
ise
d
m
od
el
helps
to
na
rrow
dow
n
on
t
he
t
op
places
t
o
vi
sit
.
The
pro
po
s
ed
m
od
el
ge
ne
rates
places
on
the
ba
sis
of
use
rs
pr
e
fer
e
nce
s
an
d
cho
ic
e.
Cu
rr
e
nt
ly
,
the
pro
po
sed
st
ru
ct
ur
e
i
s
na
rro
wed
do
wn
to
one
E
uro
pea
n
ci
ty
w
hich
ca
n
be
f
ur
t
her
util
ise
d
to
ac
hi
eve th
e
sam
e am
on
g
the
othe
r
cit
ie
s.
REFERE
NCE
S
[1]
J.
Shen,
J.
She
n,
T.
Mei
,
X.
Gao,
"La
ndm
ar
k
Rera
nking
fo
r
Sm
art
Tra
ve
l
Guide
S
y
st
ems
by
Com
bini
n
g
and
Anal
y
z
ing
Diver
se
Media
,
"
IEE
E
Tr
ansacti
ons
on
Syst
e
ms
Man
and
Cybe
rnet
ic
s:
Syst
ems
,
vol.
46(11
),
pp.
1492
-
1504
,
2016
.
[2]
Sfenria
nto
S
.
,
Sa
rag
ih
M
.
H
.
,
Nu
gra
ha
B
.
,
"
E
-
Co
mm
erc
e
Recom
mende
r
For
Us
age
Bandw
idt
h
H
ote
l
,
"
Int
ernati
o
nal
Journal
of
Elec
t
rical
and
Computer
Eng
ine
ering
(
IJE
CE)
,
vol.
9(
1),
pp
.
227
-
233
,
2017
.
[3]
G
Ado
m
avi
ci
us,
A.
Tuz
hil
in
,
A.
Tuz
hil
in
,
"Toward
the
Next
Ge
ner
ation
of
Recom
m
ende
r
Sy
ste
m
s:
A
Survey
of
the
Stat
e
-
of
-
the
-
Art
and
Pos
sible
Ext
ensions
,
"
IE
EE
Tr
ansacti
ons
on
Knowle
dge
and
Data
Engi
neer
ing
,
vol.
17
(
6
),
pp.
734
-
749
,
20
05
.
[4]
Robin
Burke
,
“
H
y
brid
Rec
om
m
ende
r
S
y
st
ems
:
Surve
y
and
Expe
riments
,
”
Us
er
Mode
li
ng
and
Us
er
-
Adapted
Inte
ract
ion
,
vol.
12(
4)
,
pp
.
331
–
3
70
,
2002
.
[5]
Vee
na
Ch,
Bahu
B.
V,
“
A
Us
er
-
Based
with
a
Scal
ab
le
Mac
hin
e
Le
arn
ing
,
”
Inte
r
nati
onal
Journal
of
El
ectric
al
an
d
Computer
Engi
n
ee
ring
(
IJE
C
E)
,
vol.
5(5)
,
pp
.
11
53
-
1157
,
2015
.
[6]
K.
Mee
han
,
T.
Lunney
,
K.
C
urra
n,
A.
McC
aughey
,
"Cont
e
xt
-
awa
re
int
e
ll
i
gent
re
comm
enda
ti
on
s
y
s
te
m
for
touri
sm
",
IEEE
Pe
rCom
,
vol
.
17
,
pp
.
328
-
331
,
2
013
.
[7]
H.
Ding,
L.
Ke,
Z.
Geng,
"Route
pla
nning
in
a
new
touri
st
rec
o
m
m
ende
r
sy
ste
m
:
A
fire
works
al
gorit
hm
-
base
d
appr
oac
h"
,
E
volutionary
Comput
ati
on
(
CEC)
2016
IEEE
Congres
s
on
,
2016
,
pp
.
4
022
-
4028.
[8]
Tobi
as
Berka
an
d
Manue
la
Pl
ößnig,
Salz
burg
R
ese
arc
h
Jakob
,
“
Designing
Rec
o
m
m
ende
r
Sy
ste
m
s
for
Touri
sm
”.
Salzburg
Re
sear
ch,
pp
.
1
-
8
,
2003
.
[9]
K.
Oku,
K.
Uen
o,
F.
Ha
tt
ori
,
"
Mapping
geotag
ged
twe
et
s
to
to
urist
spots
for
re
comm
ende
r
s
y
st
ems
,"
Proce
ed
in
gs
of
th
e
II
AI
3rd
I
nte
rnational
Co
nfe
renc
e
on
Advance
d
Appl
i
ed
I
nformatic
s
,
2014
,
pp
.
789
-
794
.
[10]
S.
Ti
wari
and
S.
Kaushik,
“
Inform
at
ion
enr
ic
hm
ent
for
touri
st
spot
rec
om
m
ende
r
s
y
stem
using
loc
ation
awa
r
e
cro
wds
ourc
ing,
”
in
Proc.
IE
EE
1
5th
Int
.
Con
f.
M
obil
e
Data
Man
age
,
2014
,
pp
.
11
–
14.
[11]
Alm
ohsen
KA,
Al
-
Jobor
H.
,
“
R
ec
om
m
ende
r
S
y
stems
in
Li
gh
t
of
Big
Da
ta
,
“
Inter
nati
onal
Journal
of
Elec
tric
al
an
d
Computer
Engi
n
ee
ring
(
IJE
C
E)
,
vol.
5(6)
,
pp
.
15
53
-
1563
,
2015
.
[12]
Pai
Chet
Ng,
Ja
m
es
She,
Ming
Cheung
and
Ale
xande
r
Cebu
ll
a
,
“
An
I
m
age
s
-
Te
x
tua
l
H
y
brid
R
ecom
m
ende
r
Sy
st
e
m
For
Vac
ation
R
e
nta
l
,
”
2016
IE
E
E
Se
cond
In
te
rn
ati
onal
Con
fe
ren
ce
on
Mul
ti
medi
a
Big
Data
,
201
6
,
pp
.
60
-
63
.
[13]
Yuan
-
Tse
Yu,
a
nd
Chung
-
Ming
Huang,
Yun
-
Tz
L
ee,
“
An
Intelli
gen
t
Tou
ring
S
y
stems
base
d
on
Mobile
Soc
i
a
l
Network
and
Cloud
Com
puti
ng
for
Tra
vel
Re
co
m
m
enda
ti
on
,
”
2
014
28th
Inte
rnational
Confe
re
nce
on
Adv
anc
e
d
Information
Ne
t
working
and
Ap
pli
cations
Works
hops
,
2014
,
pp.
19
–
24.
[14]
Kw
an
Hui
Lim
,
”
Personalize
d
R
ec
om
m
enda
ti
on
of
Tra
vel
It
ine
ra
rie
s
base
d
on
Touri
st
Inte
rests
an
d
Prefe
ren
ce
s
,
”
Jul
y
2016
In
te
rn
ati
onal
Con
fe
ren
ce
on
Us
er
Mod
el
li
ng
,
Adapt
at
io
n,
and
Pe
rs
onaliz
ati
on
,
jul
2016
.
[15]
Hari
ri
N.
,
Mob
ashe
r
B.
,
Burke
R.
,
Zhe
ng
Y.,
“
Conte
xt
-
awa
r
e
rec
om
m
enda
ti
o
n
base
d
on
r
eview
m
ini
ng
,
”
In
Proce
ed
ings
of
t
he
9th
Workshop
on
Inte
lligent
Techni
ques
for
We
b
Pe
rs
onali
z
ati
on
and
Reco
mm
ende
r
Syste
ms
in
IJCA
I
’
11
,
2011,
pp.
30
–
36
.
[16]
G.S.
Fang,
S.
K
amei,
S.
Fujit
a
,
"A
utomati
c
Gene
ration
of
Te
m
pora
l
Feat
ur
e
Vec
tors
with
Applic
at
ion
to
Touri
sm
Rec
om
m
ende
r
Sy
stems
,
"
Pro
c.
4
th
Int
.
S
ymp.
On
Computing
and
Net
working
(
CANDAR
)
,
2016,
p
p.
676
-
680
.
[17]
Chu
ghta
i
M
W,
Sela
m
at
A,
Ghani
I.
,
“
Goal
-
base
d
H
y
brid
Fi
lt
e
rin
g
for
Us
er
-
to
-
user
Personaliz
ed
Rec
om
m
enda
ti
o
n
,
”
In
te
rnationa
l
Journal
of
E
le
c
t
rical
and
Compu
te
r
Engi
n
ee
ring
(
IJE
CE)
,
vol.
3(
3)
,
pp.
329~336
,
2013
.
[18]
Go
Hira
kawa
,
G
oshi
Sato,
Kenj
i
Hisaz
um
i,
Yos
hit
aka
Shib
at
a
,
"D
at
a
g
at
he
ring
s
y
s
te
m
for
rec
om
m
ende
r
s
y
s
te
m
in
touri
sm
,
"
18th
I
nte
rnational
Co
nfe
renc
e
on
Networ
k
-
Based
Info
rm
ati
on
Syste
m
Syste
ms
,
2015
,
p
p.
521
-
525
.
[19]
Ga
y
at
re
e
Ganu
,
No
é
m
ie
E
lha
d
ad
and
Am
é
lie
Maria
n,
“
Be
y
on
d
the
Stars:
Im
proving
Ra
ti
ng
Predic
ti
ons
usin
g
Revi
ew
Te
xt
Co
nte
nt
,
”
We
bDB
,
pp.
1
-
6
,
2009
.
[20]
van
Setten
M.,
Pokrae
v
S.,
Ko
olwaa
i
j
J.
,
“
Con
te
xt
-
Aw
are
Recom
m
enda
ti
ons
in
the
Mobil
e
T
ourist
Application
COM
PASS
”.
In
De
Bra
P.
M
.
E
.
,
N
ej
d
l
W.
(
eds)
Inte
rnat
ional
C
onfe
renc
e
on
Ad
apti
v
e
Hyp
ermedi
a
and
Adapt
i
v
e
We
b
-
Based
S
yst
ems.
AH
2004
.
L
ec
ture
Note
s
in
Computer
Scien
ce
,
vol
.
3137
,
pp
.
235
-
244
,
2004
.
[21]
S
Girish
,
B
Ram
amurth
y
,
T
Sent
hil
nat
h
an
,
“
Mining
the
W
eb
Da
ta
for
Cla
ss
if
y
ing
and
Predicting
U
sers’
Reque
sts
,
”
Inte
rnational
Jo
urnal
of
El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
,
vol
.
8(4)
,
pp.
2390~2398
,
2018
.
Evaluation Warning : The document was created with Spire.PDF for Python.